On the distance to the giant component along a straight line in a two dimensional percolation
5
pages
English
Documents
Obtenez un accès à la bibliothèque pour le consulter en ligne En savoir plus
Découvre YouScribe et accède à tout notre catalogue !
Découvre YouScribe et accède à tout notre catalogue !
5
pages
English
Documents
Obtenez un accès à la bibliothèque pour le consulter en ligne En savoir plus
On the distance to the giant component along a straight line in a two-dimensional percolation model. Olivier Dousse School of Computer and Communication Sciences EPFL CH-1015 Lausanne, Switzerland Christina Tavoularis School of Electrical and Computer Engineering Cornell University Ithaca, NY, USA Patrick Thiran School of Computer and Communication Sciences EPFL CH-1015 Lausanne, Switzerland Abstract— The supercritical regime of a percolation model refers to the range of probabilities (discrete) or densities (con- tinuous) above a critical value for which there exists a unique unbounded cluster almost surely. In this paper, we provide an upper bound to the linear distance from the origin to this giant connected component for both the discrete and the continuous (Boolean) model in two-dimensions. By modeling a dense wireless sensor network with a supercritical Boolean model, our result bounds the distance traveled by a target moving in a straight line before it is detected by a node who can relay the alert through a multihop path to the sink. This result incorporates a solidified definition of detection requiring that the intrusion alert successfully reach the central authority. I. INTRODUCTION Consider a wireless multihop sensor network whose task is to detect the intrusion of a moving object in the area in which the sensors are deployed.
Voir
- until detection
- look only
- all clusters
- when
- infinite cluster
- exist finite constants
- dimensional percolation
Publié par
Langue
English
On the distance straight line in
Olivier Dousse School of Computer and Communication Sciences EPFL CH-1015 Lausanne, Switzerland olivier.dousse@epfl.ch
to the giant component along a a two-dimensional percolation model.
Christina Tavoularis School of Electrical and Computer Engineering Cornell University Ithaca, NY, USA crt23@ece.cornell.edu
Abstractsupercritical regime of a percolation model— The refers to the range of probabilities (discrete) or densities (con-tinuous) above a critical value for which there exists a unique unbounded cluster almost surely. In this paper, we provide an upper bound to the linear distance from the origin to this giant connected component for both the discrete and the continuous (Boolean) model in two-dimensions. By modeling a dense wireless sensor network with a supercritical Boolean model, our result bounds the distance traveled by a target moving in a straight line before it is detected by a node who can relay the alert through a multihop path to the sink. This result incorporates a solidified definition of detection requiring that the intrusion alert successfully reach the central authority.
I. INTRODUCTION Consider a wireless multihop sensor network whose task is to detect the intrusion of a moving object in the area in which the sensors are deployed. The first question that comes to mind is to know how far the object can move in the monitored area before being detected by a sensor. When the object moves along a straight line, stochastic geometry provides the answer under the name oflinear contact (or hit) distribution function (see e.g. [1], p. 80). For example, when the sensor spatial distribution is Poisson, this function decreases exponentially with the distance traveled by the object. The sensor that has detected the intruder needs however to convey this information to a monitoring station that collects the sensed data, thesink. Nodes between the detecting sensor and the sink act as relays for the message. However, various sources of noise, battery failures in the nodes or simply their random locations, will inevitably result in having some sensors disconnected from the network. If the sensor that detected the intruder is not connected to the sink, the intruder can continue to progress in the monitored area without actually being detected by the central sink. Consequently, a second, more difficult question is to know the distance that can be
The work presented in this paper was supported (in part) by the National Competence Center in Research on Mobile Information and Communication Systems (NCCR-MICS), a center supported by the Swiss National Science Foundation under grant number 5005-67322.
Patrick Thiran School of Computer and Communication Sciences EPFL CH-1015 Lausanne, Switzerland patrick.thiran@epfl.ch
traveled by the intruder until the sink (and not only any sensor) can be notified. This paper addresses precisely this question. We assume that the sensor network can be described by a 2-dimensional Poisson Boolean model: nodes (sensors) are distributed according to a 2-dimensional homogeneous Poisson process of intensityλ, their sensing range is identical to their transmission range, and is randomly distributed, independently from all other nodes. In such a model, it is well known (see. e.g. [2]) that given the distribution of the ranges, there is a critical intensityλc>0such that all clusters of connected nodes are almost surely finite ifλ < λc(subcritical phase) and a single giant, unbounded cluster of connected nodes appears almost surely ifλ > λc(supercritical phase). We assume in this paper thatλ > λc, and that the sink belongs to the giant cluster. Suppose that the intruder starts at the origin and moves along a straight line in any arbitrary direction. How far can it move until it hits the giant cluster? After having briefly reviewed the state of the art in Section II, we answer this question first in a discrete bond percolation setting in Sec-tion III. Using this construction, we then answer the question in the Boolean continuous model in Section IV. Section V concludes the paper. II. RELATED WORK Tracking and detecting a moving object is an important application of a sensor network, and has thus received some attention. Most of the work is so far devoted to the problem of computing the linear contact (or hit) distribution function, i.e. the distance traveled by the intruder until detection by a sensor without checking that the sensor that spawns the alarm is actually connected to the network. The simplest case is to compute this distance when the network is modeled by a Boolean model; it is known to be exponentially distributed (see e.g. [1], p.80). More results from this approach are derived in [3]. When the motion of the object does not follow a straight line, but a Brownian motion, explicit formulas and/or bounds are obtained in [4]. To save energy, nodes periodically switch
